[gmx-users] Constructing New Metal-Containing Topologies in OPLS

[M. Jake Pushie]

Hello gmx'ers,

I've been developing new residue topologies for the OPLS FF. These are
backbone-bound Cu-complexes, where the metal is coordinated directly to a
deprotonated amide N-atom, such as is observed at the N-terminal end of
serum albumin. I’m getting reasonable results, but unfortunately reviewers
seem to have had no end of complaints about the way we have represented
these things, so I'm hoping some experts out there could offer some useful
suggestions on how I might proceed with the modeling, or at least hint
that I really am on the right track.

Here's our current procedure:
1. Starting geometry for backbone atoms and the metal centre from X'tal
structure
2. Harmonic potentials for Cu-ligand bonds
3. Harmonic potential for the backbone atoms immediately around the metal
centre (these are planar complexes)
4. 1-4 pairs list updated to include the copper 1-4 interactions in the
models
5. CHELPG charges calculated using B3LYP/6-31G(d) from an optimized X'tal
geometry of the metal coordination site (at that same level of theory)

The main complaints of our models have been:
(1) the CHELPG charges that come out of the calculations are too small and
would favour hydrophobic-types of interactions. One reviewer seems to be
suggesting we start out with ion-like charges on the metal and ligands and
go from there. I don't really like this idea as it's not consistent with
the way the charges are derived for OPLS. Also any other charge-derivation
method would no longer be consistent with OPLS(!). Comparing the backbone
and side chain charges I calculate with their related atoms in similar
molecule fragments in the OPLS force field they differ by less than 5-10%
in most cases, except those directly involved in metal-coordination (which
is not unexpected).
(2) putting a harmonic potential on backbone dihedrals about the metal
centre makes these things too flat. I say: Since Cu binds into the
backbone, it replaces an H-atom and therefore the dihedral angle from Cu
across the backbone to the carbonyl O-atom is actually somewhere between
160-180 degrees, and without explicitly entering the local dihedral angles
based on X'tal data the force field only has the Cu-ligand bonds and
backbone angles holding it in place, but nothing else to define local
coordination geometry and backbone orientation.

Does anyone who has done CHELPG charge derivation have suggestions on the
CHELPG charge comments above? I'm using a radius of 1.4 A. for Cu2+. The
complexes are overall charge-neutral, and the CHELPG charge on Cu usually
comes out to be between 0.6 to 0.8 for all of the related complexes I've
calculated (I have data for more than 20 related species in various
protonation states and with different amino acids).  This apparently low
charge density (instead of something closer to 2+) at Cu may be because it
is bound by two formally anionic backbone amide N-atoms that are
deprotonated.

Also, I have recently run B3LYP/6-31G(d) geometry optimizations and
frequency calculations for a large number of hydrated Cu2+ complexes with
various ligands, and am using the force constants from the frequency
calculations (in mDyne/Angstrom, scaled by 60230) in the harmonic bond
potentials for the copper ligands. Does anyone know if it is reasonable to
expect an appreciable change in the magnitude of the force constant that
is calculated depending on the overall charge metal complexes (i.e. 1+,
2+, neutral, etc.)?

Jake Pushie

Ph.D. Candidate
Structural Biology Research Group
University of Calgary
Calgary, AB
CANADA